What Are OpenAI dots in 2026?
Dots are always-on AI agents OpenAI announced on 29 September 2026, powered by GPT-6 Astra. Each dot has its own cloud computer and its own browser, learns from your feedback over time, and can work towards your goals around the clock. Through OpenAI's plugin ecosystem they connect to more than 4,000 apps, and you reach them in ChatGPT on desktop, web and mobile, or in Slack and Teams, with texting said to be coming.
The distinction from everything before it is persistence. ChatGPT Work runs a task when you ask. A dot keeps existing. You name it, it learns your preferences and standards, it messages you with progress or questions, and when you are not actively working with it, it looks for ways to help. OpenAI calls that last behaviour proactive research, and it is the part with the most interesting constraints attached.
Explain It Like I Am Five
Imagine you hire an assistant who sits at their own desk, with their own computer, in their own room. You can call them, message them, or drop in to watch what they are doing. They remember how you like things done, so next time you do not have to explain. When you are busy they quietly look around for useful things to do, but while they are looking they are only allowed to read, never to touch anything. And for the important stuff, sending money, changing a password, they must come and ask you first. That is a dot: not a tool you pick up and put down, but someone at a desk who is always there.
| Attribute | What OpenAI published in 2026 | Why a business team should care |
|---|---|---|
| Announced | 29 September 2026, rolling out in ChatGPT | Pro and Business Premium today, Enterprise beta by admin switch. |
| Model | Powered by GPT-6 Astra | The computer-use flagship, which is why a dot can operate software. |
| Environment | Its own cloud computer and browser, inspectable at any time | Your machine stays separate unless you connect it. |
| Integrations | Over 4,000 apps via plugins | Existing ChatGPT app controls govern access. |
| Where it lives | ChatGPT desktop, web, mobile, plus Slack and Teams | Context carries across every channel. |
| Cost | First dot included in Pro or Business Premium at no extra cost | Plan includes an allowance for deeper work, with extended limits for the first month after launch. |
| Usage | Conversations with your dot do not count toward ChatGPT usage limits | But tasks it starts in Codex or ChatGPT Work do count as usual. |
| Setup | Create your first dot in the desktop app or a desktop browser | Mobile messaging works after initial setup, not before. |
Why Does an Always-On Agent Matter for Business Teams in 2026?
Because it removes the trigger. Every AI workflow until now has needed a human to start it, which means the work only happens when someone remembers. A dot that is already running changes which jobs are worth delegating: the small recurring things nobody gets around to, the follow-ups that slip, the monitoring that only matters when something changes.
OpenAI's own examples are deliberately mundane, which is the point. A bug appears in Slack and dots start investigating. A design arrives and dots turn it into a working app while the team handles customer feedback. A planning cycle starts and dots keep everyone moving toward the deadline. The one from an external tester is the most telling: his dot noticed he had forgotten to invoice a publication, prepared the invoice, and sent it after his approval. Nobody asked it to check.
OpenAI writes that the magic of dots is when they bring you work done the way you would do it, sometimes before you even think to ask. That is a genuinely different product promise, and it is also the reason the permission model below deserves more of your attention than the capability list. An agent that acts before being asked needs boundaries that hold when you are not watching.
How Do dots Work on Their Own in 2026?
Each dot operates from its own cloud computer with its own browser and whichever apps you have connected. You can open that computer at any time to inspect the work, which is a meaningful transparency feature rather than a dashboard summary. A dot can take a project and run with it while several others are in flight, so you keep handing it work without managing separate threads.
It can also reach beyond its own machine. Dots can connect to other devices, and you can give a dot permission to use your laptop, where it works directly alongside you. That is opt-in, and it is the boundary most worth thinking about before you grant it, because it moves the agent from a sandbox into your actual environment.
What Are the Safeguards on dots in 2026?
OpenAI published a specific and unusually legible permission model, and the most important constraint sits on the proactive behaviour. When you are not actively working with it, your dot looks for ways to help using the apps you have already connected, but with tools restricted to read-only. OpenAI states plainly that during proactive research a dot cannot send messages, change app content, or control your browser or computer.
- Separation by default: each dot works on its own cloud computer. Your computer and its contents stay separate unless you choose to connect them.
- Credentials without exposure: for signing into supported websites, dots can use saved passwords without exposing them to the model.
- Read-only when unattended: proactive research is restricted to read-only tools, as above.
- Monitoring that can stop work: security safeguards defend against malicious instructions and monitor for potentially harmful behaviour. If monitoring detects a safety concern it can pause or stop the dot's work.
- Custom Rules: dots start with built-in rules for when to act independently and when to ask. Custom Rules let you allow specific actions, require approval, or block them, and built-in safety requirements always apply.
- Auto-review on consequential actions: dots use auto-review to check actions that could affect your accounts or share information against your instructions, Custom Rules and safety requirements, deciding what proceeds, what needs approval and what you must do yourself.
- Some things never delegate: OpenAI names changing a password as an example of a sensitive task that always stays with you.
- Visible progress: Activity View shows progress including background work, and you can redirect a dot as needed.
OpenAI also closes with the sentence every team should paste into its own policy: dots can still make mistakes, so always review consequential work.
What Happens to Your Data With dots in 2026?
The training position differs by plan, and the detail about proactive research is worth knowing. OpenAI states it does not use content from ChatGPT Business, Enterprise or Edu workspaces to improve its models by default. On personal ChatGPT plans you control whether your dots' conversations and work are used to improve models.
Specifically on the always-on behaviour, OpenAI says it does not train directly on proactive research or on your dot's notes to itself, but that information from them may be used if it helps inform an eligible conversation or task, depending on your settings. That is a narrower carve-out than "we do not train on it", and if you handle client data, that is the sentence to take to your own privacy review rather than the headline.
Who Can Use dots, and What Does It Cost in 2026?
Dots are rolling out in ChatGPT to Pro and Business Premium users in eligible markets. Enterprise users, including Edu and Healthcare, can try the beta once a workspace admin enables it. OpenAI says it plans to expand to more users soon.
| Plan | Access in 2026 | What to do |
|---|---|---|
| Pro | Rolling out now in eligible markets | First dot included at no extra cost. |
| Business Premium | Rolling out now in eligible markets | First dot included at no extra cost. |
| Enterprise, Edu, Healthcare | Beta, once the workspace admin enables it | Admin action required before anyone sees it. |
| Other plans | Not yet | OpenAI says expansion to more users is planned. |
On economics, the structure matters more than the number. Your first dot is included, and your plan includes an allowance for deeper work with extended limits for the first month after launch. Conversations with your dot do not count toward ChatGPT usage limits, but when you ask your dot to start or manage tasks in Codex or ChatGPT Work, those tasks count toward your usage limits as usual. OpenAI says you will later be able to add more dots and scale each one by increasing its speed or its total monthly work.
Read that pairing carefully before budgeting. Talking to a dot is free of your limits; making it do the heavy work is not. A team that measures dot adoption by conversation volume will conclude it is free, right up to the point where the delegated Work and Codex tasks land on the allowance.
What Are Specialist dots, and Why Do They Matter in 2026?
Specialist dots take on dedicated responsibilities inside an organisation rather than working for one person. Your company sets each one up with its own identity, credentials and access to the systems it needs, and OpenAI says the dot adapts to new information and improves with feedback from your team. This is previewed rather than generally available, starting with focused enterprise pilots where OpenAI's engineering teams work directly with organisations to define each dot's responsibilities, its tools, and how people review and approve its work.
The functions OpenAI names from its own early internal testing are the tell: procurement, invoice processing, email marketing, customer support and commercial contracting. Email marketing sitting in that list is the one marketing leaders should notice, because it signals where OpenAI expects a scoped, credentialed agent to own a recurring function rather than assist with it.
OpenAI also says it is working with Microsoft to integrate specialist dots with enterprise governance and security controls in Agent 365, so businesses can manage dots through Microsoft tools they already use. For anyone whose IT function standardises on Microsoft, that is the sentence that decides whether this becomes procurable.
How Should Marketing and Growth Teams Use dots in 2026?
Point a dot at the recurring work that slips because nobody owns the trigger, and keep the approval gate on anything that reaches a customer or an account. The read-only constraint on proactive research is what makes unattended monitoring reasonable, and the auto-review layer is what should make you comfortable letting it prepare an action it cannot complete alone.
| Workflow | Fit for a dot in 2026 | Control to set |
|---|---|---|
| Monitoring a client Slack channel for escalations | Strong. Read-only proactive research suits watching. | Let it flag to a human. No outbound messages. |
| Recurring reporting assembled from connected apps | Strong, and it can start before you ask. | Approval before anything is shared externally. |
| Follow-ups and invoices that slip | Strong. This is OpenAI's own tester example. | Require approval on send. Never auto-send. |
| Campaign launch readiness checks | Good. It can inspect and prepare. | Human sign-off on tracking, claims and consent. |
| Competitive and market monitoring | Strong fit for proactive research. | Verify sources; treat output as a lead, not a fact. |
| Anything that emails a customer or changes an ad budget | Only behind Custom Rules requiring approval. | Block or require approval explicitly. Do not rely on defaults. |
| Credential or account changes | Not delegable. OpenAI keeps password changes with you. | None needed. It stays yours. |
One setup habit is worth building in 2026: write your Custom Rules before you connect your apps, not after. The defaults are OpenAI's judgement about when to act independently, and they are reasonable, but they are not your risk tolerance or your client's. The agency that decides its approval boundaries on day one will not have to discover them through an incident.
What Are the Common Mistakes to Avoid With dots in 2026?
- Assuming it is free because chat is free. Conversations do not count toward usage limits. Delegated Codex and ChatGPT Work tasks do.
- Missing the first-month limits. Extended limits apply for the first month after launch, so early usage is not a reliable guide to steady-state cost.
- Connecting your laptop by reflex. Each dot is separate by default. Granting access to your own machine is a deliberate, reversible-with-effort decision.
- Treating proactive research as fully passive. It is read-only while unattended, which is strong, but information from it may inform eligible tasks depending on your settings.
- Leaving the default rules in place for client work. Custom Rules exist. Use them to require approval on anything outbound.
- Expecting specialist dots now. Those are a preview with focused enterprise pilots, not a self-serve feature.
- Skipping review because auto-review exists. OpenAI's own closing line is that dots can still make mistakes, so always review consequential work.
Key Takeaways for 2026
Dots move AI from a tool you invoke to a colleague who is already working, and the interesting part of the launch is the permission model rather than the capability list.
- Announced 29 September 2026, powered by GPT-6 Astra, each with its own cloud computer and browser and access to over 4,000 apps through plugins.
- Rolling out to Pro and Business Premium in eligible markets; Enterprise, Edu and Healthcare by admin-enabled beta.
- First dot included at no extra cost, with an allowance for deeper work and extended limits for the first month after launch.
- Conversations do not count toward ChatGPT usage limits, but tasks started in Codex or ChatGPT Work do.
- Proactive research is restricted to read-only tools: no sending messages, no changing app content, no controlling your browser or computer.
- Auto-review, Custom Rules and Activity View are the control surface, and password changes always stay with you.
- Specialist dots with their own identity and credentials are a preview via enterprise pilots, with Microsoft Agent 365 integration in progress.
- OpenAI's own guidance: dots can still make mistakes, so always review consequential work.
Distk helps growth teams across India and internationally decide which recurring work is safe to hand to an always-on agent, write the approval rules before the apps get connected, and keep the human checkpoint on everything that touches a customer or a budget. If dots are arriving in your workspace in 2026, that design is where we start.
Sources
- OpenAI, Introducing dots, 29 September 2026. Every capability, permission, plan and billing detail in this guide comes from that announcement and is OpenAI's own documentation.